Biologically-Informed Hybrid Membership Inference Attacks on Generative Genomic Models

The increased availability of genetic data has transformed genomics research, but raised many privacy concerns regarding its handling due to its sensitive nature. This work explores the use of language models (LMs) for the generation of synthetic genetic mutation profiles, leveraging differential privacy (DP) for the protection of sensitive genetic data. We empirically evaluate the privacy guarantees of our DP modes by introducing a novel Biologically-Informed Hybrid Membership Inference Attack (biHMIA), which combines traditional black box MIA with contextual genomics metrics for enhanced attack power. Our experiments show that both small and large transformer GPT-like models are viable synthetic variant generators for small-scale genomics, and that our hybrid attack leads, on average, to higher adversarial success compared to traditional metric-based MIAs.

Key Contributions

• Novel biHMIA attack that fuses traditional black-box MIA with biologically-informed genomic metrics (e.g., allele frequency, SNP rarity) for enhanced attack power
• Empirical evaluation showing biHMIA achieves higher average adversarial success than traditional metric-based MIAs against DP-protected generative genomic models
• Demonstration that small and large GPT-like transformer models are viable synthetic genetic mutation profile generators for small-scale genomics settings

🛡️ Threat Analysis

Details

Similar Papers